Work Experience

Agent-based Modeling | Transport Modeling | Spatial Analytics

Jayita Chakraborty, PhD
Data Scientist | University of Leeds
Jan 12, 2026

Research Project,
School of Geography, University of Leeds

Analysing Global Policy Responses to COVID-19

Background

  • Studying rapid decisions to inform future responses to similar crises.
  • Identify gaps in preparedness and areas for improvement.
  • Evaluate the effectiveness of different strategies and inform the development of policies and procedures to better handle future emergencies.

RQ: What mechanisms drive the rapid and homogeneous diffusion of lockdown policies across heterogeneous countries, and how can these mechanisms be captured through agent-based simulation?

Methodology

Attributes Characteristics
GDP National income per country that is read from data
Democracy Democracy index per country that is read from data
Population Density Average Population density per country that is read from data
Covid Cases Covid Cases per thousand population (normalized on average)
Initial State of policy Adopted policy measure on March 1, 2020
Distance (Minimum Difference) Key variable of the model in that defines the distance between some agent and all other agents
Social Threshold A parameter incorporated to influence the policy adoption of the agent. The parameter is calculated based on the following equation: \[ ST = e^{\text{Covid Cases}} \times (\text{Base Threshold})^{2} \times \text{Lockdown Adoption Probability} \]

Base threshold:

  • A parameter introduced to affect the social threshold value of the countries
  • Ranges from 0.02 to 0.25
  • Optimal value 0.13

Peer size:

  • The number of countries effective in influencing policy decision.
  • Ranges from 10 to 25.
  • Optimal value 18.

Lockdown adoption probability:

  • The probability of countries to adopt a lockdown on their own.
  • Ranges from 0.0002 to 0.07.
  • Optimal value 0.01.

Outputs

PhD Research Project,
IIT, Kharagpur & Curtin University

A Simulation-based assessment of demand-supply interactions for Ridesourcing services in an urban environment

Background

  • Demand-supply imbalance
  • Increases overall VKTs of ridesourcing vehicles compared to the passenger using his or her personal vehicle to conduct the same trip
  • Competition with other motorised modes
  • The operation of the TNCs are yet to have a clear picture before the regulators, making it difficult for the regulators to define these services within suitable legal framework (Flores & Rayle, 2017).

Research Question

  • Which operational parameters are most critical for evaluating the performance and efficiency of ridesourcing systems?
  • How do driver behavioural characteristics influence the dynamics of ridesourcing operations? What are the threshold values for driver idle waiting time after completing a trip? What strategies do drivers adopt for passenger search once their waiting time tolerance is exceeded?
  • How do user behavioural traits affect their tolerance for waiting in ridesourcing systems? What is the threshold limit for user waiting time?
  • How do different passenger search strategies and ride-matching algorithms affect system performance in simulated scenarios?

Research Project,
ITS, University of Leeds

Where do e-cargo bikes go?

Research Question

Challenges

  • Problematic points
  • Discrepancy between datasets
  • Missing information

Methodology

  • The bike is moving but the route points are being recorded in the wrong place
  • Records random “Dodgy” or erroneous coordinates
  • Conditions based on rolling average distance, speed to next point and time difference
  • Bike not moving; but the GPS records points in the wrong place.
  • GPS receiver tries to acquire navigation data and estimates a position.
  • Filter points based on whether the point is random compared to its neighbours
  • Issues with GPS signal; May have occurred due to GPS device not being turned on.
  • Missing information on route
  • Search and insert missing points if available
  • Bike is on the move; Interference with GPS signal.
  • Missing information on route
  • Search and insert missing points if available
  • Speed of the bikes higher than usual
  • Bikes being transported in car/train
  • Incorrect information on route
  • Update speed and identify whether the vehicle was being transported
  • Identifying trips from the cleaned GPS points
  • Identification of stop criteria
  • Tsui (2005) suggested dwell time between trips can vary from 120s to 600s.
  • Use OS Terrain 50 DEM – an open dataset that is more accurate than often used NASA SRTM data
  • Use R Extract function – adds Elevation value to the SF dataframe of route points
  • Calculate elevation gain/descent for each trip

Outputs

Before applying algorithm

After applying algorithm